Method for identifying public space social behavior and environmental impact based on a network of perception devices

Abstract

The application discloses a public space social behavior and environment influence recognition method based on a perception device network, and comprises the following steps: public space area selection, measuring point layout and perception device distribution, public space physical environment monitoring, public space social behavior recognition, behavior environment index influence analysis, influence rating and output; the application is based on the perception device network layout, monitors the public space physical environment through the environment perception device, recognizes the public space social behavior through the behavior perception device, and comprehensively analyzes the influence relationship between the environment index and the behavior index, performs influence degree sorting and output, and the application can be used for recognizing the behavior environment influence relationship in the city intelligent perception field, and helps to create more efficient and comfortable city public spaces.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of urban intelligent perception, and particularly relates to a public space social behavior and environment influence identification method based on a perception device network. BACKGROUND

[0002] Under the background of urban high-density development, open urban public spaces concentrate the carrying of most outdoor social activities, and their physical environment is directly related to the life quality of residents and the social interaction mode. Reasonable public space physical environment creation helps to promote the communication and interaction among residents, promote the contact and mutual assistance within the community, and enhance the community cohesion and sense of belonging.

[0003] At present, the monitoring methods of public space physical environment mainly include remote sensing and ground observation. Among them, the remote sensing method is affected by atmospheric interference and ground cover, and the data analysis error is large; the ground observation method is more suitable for micro urban spaces such as streets and squares, but the current technology pays more attention to the single latitude of thermal environment and wind environment, and less attention to the comprehensive measurement of physical environment. The behavior identification of public space mainly includes image recognition and inertial measurement methods. Among them, the image recognition method has less dependence on wearable devices and is more suitable for long-term monitoring of public space social behavior, but the current technology pays more attention to the application scenarios such as driving safety and indoor security, and lacks technical methods for urban social behavior identification.

[0004] Therefore, it is urgent to solve the above problems. SUMMARY

[0005] The purpose of the application is to provide a public space social behavior and environment influence identification method based on a perception device network. The application is based on a perception device network integrating environment perception devices and behavior perception devices to realize the monitoring and identification of public space social behavior and physical environment and the analysis of influence relationship.

[0006] Technical scheme: In order to achieve the above purpose, the application discloses a public space social behavior and environment influence identification method based on a perception device network, which includes the following steps:

[0007] (1) selecting a public space area through an interactive map interface of a webpage, and recording the geographic position information of the public space area based on the selection input of the user;

[0008] (2) determining the number of perception devices and the layout of measuring points based on the geographic position information of the public space area in step (1), deploying a perception device network in the public space area, the perception device network including environment perception devices and behavior perception devices, testing the working state and coverage range of the devices, and transmitting continuous monitoring data to a data center;

[0009] (3) The physical environment data of the public space area collected by the environment perception device is summarized according to the time node to the corresponding space grid unit, a grid data matrix with time sequence information is generated, and the continuous environment data distribution of the public space area is generated through the space interpolation method, the average value of the environment data in the unit time interval is calculated, and the structured data is stored in the database;

[0010] (4) The video data of the behavior perception device is processed in real time through the convolutional neural network, the individual in the crowd is detected and tracked, the individual position coordinates and motion trajectory are obtained, the influence of the shelter and the video angle is eliminated through the multi-camera data fusion technology, and the redundant data is combined, the social state and the social behavior type of the individual are judged by combining the social distance, the face orientation and the limb action recognition, the space grid data is added to the social behavior index, and the structured data is stored in the database;

[0011] (5) Based on the structured data of the physical environment in step (3) and the structured data of the social behavior in step (4), the influence of the fixed space position and the time node on the social behavior is excluded through the space regression model and the time sequence decomposition, and the social behavior data set only retaining the influence of the environmental factors is obtained; the influence of each environmental index on the social behavior is judged, and the influence mode and degree are determined by combining the correlation analysis and the multiple regression analysis;

[0012] (6) The influence degree of each environmental index of the public space on the social behavior is sorted and three-dimensionally displayed and output according to the regression analysis result in step (5).

[0013] Optionally, step (1) specifically comprises the following steps: an interactive map is loaded and displayed on a web interface through an HTML container element, user selection input on the map is received, the selection input refers to the selection of the public space area in the map; the geographic location information of the public space area is recorded based on the selection input of the user, and the geographic location information includes all vertex coordinate geometric information, geometric center point geographic coordinates and geographic area of the public space area.

[0014] Optionally, in step (2), the environment perception device integrates a temperature and humidity sensor, a light sensor and a wind environment sensor, and is evenly distributed in the public space area at a 20m x 20m grid; the behavior perception device integrates a movable power supply and a network camera, and the effective sensing range of the behavior perception device is a planar area with a radius of 50m and an angle of 80° centered on the device, the boundary and the geometric center of the public space area are determined, the public space area is divided into 5m x 5m grid units, and the number and distribution positions of the devices achieving full grid unit coverage and minimum overlap degree are calculated through a genetic algorithm.

[0015] Optionally, the structured data in step (3) includes grid cell identifiers, timestamps, environmental index types, and data average values, the environmental index types include light environment indexes, wind environment indexes, and thermal environment indexes, the light environment indexes include sunshine hours per minute and per hour, the sunshine hours are calculated according to the time when the direct radiation measured by the light sensor is less than or equal to 120 W / m2; the wind environment indexes include average wind speed per minute and average wind direction, the average wind speed per minute and the average wind direction are calculated according to the wind speed and the wind direction measured by the wind environment sensor; the thermal environment indexes include average temperature per minute and relative humidity, the average temperature per minute and the relative humidity are calculated according to the temperature and the humidity measured by the temperature and humidity sensor.

[0016] Optionally, the multi-camera data fusion technology in step (4) refers to combining the position coordinates (x c ,y c ,z c ) and the shooting direction (θ x ,θ y ,θ z ) of the camera, the center point coordinates (c x ,c y ) of the individual image in the photographic image, and the focal lengths f x and f y of the x-axis and the y-axis on the photographic image plane determined by the vertical angle of the camera and the sensor size of the camera, to construct the intrinsic matrix and the extrinsic matrix [R|T] of the camera, wherein the rotation matrix R = R z (θ z )·R y (θ y )·R x (θ x ), wherein R x , R y , R z are rotation matrices of rotation angles around the x, y, and z axes, respectively, the translation matrix wherein t x , t y , and t z are translation amounts of the center point coordinates of the individual image in the photographic image along the x, y, and z axes, respectively, and the projection matrix P = K·[R|T] of the camera is defined, and the individual image coordinates are converted into spatial plane coordinates through the projection matrix; a series of grid coordinate points are recorded based on the converted spatial plane coordinates and the trajectories of each individual in the camera, the trajectory data of each camera are matched through timestamps and spatial positions, and the distance between two trajectories in a unified time period is calculated wherein (X i,t ,Y i,t ) represents the coordinate position of individual i at time t, (X j,tY j,t ) represents the coordinate position of individual j at time frame t; set the distance threshold to 2m, and identify the individual similarity by ReID (Re-Identification) across devices when d < 2m; use the ResNet deep learning model to extract the feature vector of the individual image captured by each camera, and calculate the feature vector similarity between different individuals within a unified time period under different cameras by Cosine Similarity:

[0017]

[0018] Wherein: x·y is the dot product of the feature vector, and ||x|| and ||y|| are the Euclidean norms of vectors x and y respectively; set the cosine similarity threshold to 0.5, and consider the feature vectors with cosine similarity > 0.5 as the same object and assign the same identity ID; by associating the trajectories of the same identity ID in different cameras and merging the trajectories matched in each camera, remove the duplicate trajectories and form unified trajectory data; merge the matched trajectories in different cameras into a continuous trajectory sequence, retain the plane coordinates, video frame images and identity ID information in each time frame, and store the trajectory data including the plane grid coordinates (X t ,Y t ) and the corresponding video images of each time frame t after merging into a structured data form to the database.

[0019] Optionally, in step (4), the social state and social behavior type of the individual are judged by combining social distance, face orientation and limb action recognition. First, the database is sorted, the motion trajectory of the individual is segmented according to the time stamp, and the video frame is extracted at 0.5s time interval; by analyzing the motion trajectory, according to the distance between any two individuals at different time points of the plane coordinate data, for each time stamp t, the Euclidean distance d ij (t) between individual i and individual j is calculated:

[0020]

[0021] Wherein: x i (t), y i (t) is the plane coordinate of individual i at time t, based on the preset social distance threshold D threshold , if d ij (t) ≤ D threshold and the duration Δt exceeds the threshold, it is judged that there is social interaction between two or more individuals; by positioning the face region through the MTCNN face detection algorithm, calculating the Euler angle (α, β, γ) of the head through head pose estimation, matching the facing direction between individuals, if the yaw angle βi and β j the difference between β threshold is within a threshold Δβ i , i.e., |β j |≤Δβ threshold , and the relative distance meets the social distance threshold, it is determined that it is conducting a social interaction; the key node coordinates P={p1, p2...p n} in each video frame are extracted by the OpenPose human pose estimation algorithm, the motion trajectory and speed of the key limb parts are calculated by analyzing the time series of the key nodes, and the movement distance of the key node k between time t and t+Δt:

[0022]

[0023] The extracted feature vector v={v1, v2...v n} is input into the SVM classification model, and the social distance d ij (t), the face orientation β, and the limb motion feature v are comprehensively analyzed:

[0024] F=concat(d ij (t), β i , β j , v)

[0025] The probability distribution P(class|F) is output by the random forest model, and the feature vector F obtained by splicing multiple features is calculated to correspond to the three behavior types of health class, entertainment class, and ordinary social class, and finally a structured data set including individual ID, motion trajectory, social state, social behavior type, time and space position of behavior occurrence, and corresponding motion state detailed information is output.

[0026] Optionally, the structured data in step (4) includes grid cell identifiers, timestamps, behavior indicator types, and indicator data values, the behavior indicator types include social state and social behavior type, the social state includes two data values of yes and no, and the social behavior type includes three data values of health class, entertainment class, and ordinary social class.

[0027] Optionally, in step (5), the influence of fixed spatial position and time node on social behavior is excluded by a spatial regression model and time series decomposition. First, the spatial autocorrelation of the structured social behavior data is calculated, a spatial weight matrix W is constructed for capturing the mutual relationship between spatial positions, wherein the element W ij of the matrix represents the spatial interaction between position i and position j, and a spatial autoregressive model SAR is constructed:

[0028] Y ij =ρWY ij+X ij β+∈ ij

[0029] where Y ij is the social behavior of individual i at location j, ρ is the spatial autocorrelation coefficient, X ij is the environmental index, β is the coefficient vector of the independent variables, and ∈ ij is the error term;

[0030] The model parameters ρ and β are estimated using maximum likelihood estimation (MLE), and based on the estimated ρ, the spatial autocorrelation part of the social behavior data is calculated:

[0031] Y spatial = ρWY

[0032] The spatial autocorrelation part is removed from the original data to obtain the non-spatial autocorrelation part, i.e., the residual Y residual :

[0033] Y residual = Y - Y spatial - Xβ

[0034] The residual Y residual is the social behavior data after removing the spatial autocorrelation and excluding the fixed spatial position;

[0035] Secondly, the structured social behavior data is decomposed into a time series to calculate:

[0036] Y t = T t + S t + R t

[0037] where Y t is the social behavior data, T t is the trend component, S t is the seasonal component, and R t is the residual component; the trend and seasonality are removed from the original data to obtain the residual R t = Y t - T t - S t , which is the social behavior data excluding the influence of time nodes; the residual data Y residual from the spatial autocorrelation model is combined with the residual data R t from the time series decomposition to obtain the comprehensive spatial and temporal adjusted social behavior data set Y adj :

[0038] Y adj = Y residual + R t

[0039] Dataset Y adj The influence of fixed spatial position and time node on social behavior is excluded, and the characteristics of behavior change related to environment are retained.

[0040] Optionally, in step (5), the influence of each environmental index on social behavior is determined by combining correlation analysis and multiple regression analysis, and the influence mode and degree are determined. That is, the environmental index data in the structured data of the physical environment in step (3) is standardized, and the social state and three behavior type data in dataset Y adj in step (1) are all coded into binary variables, where "0" represents "no" and "1" represents "yes"; the correlation coefficients of the environmental index data and the social state data and the three behavior type data are calculated by point two column correlation analysis, and the environmental indexes significantly correlated with the social state and the behavior type are screened out, where the significant correlation refers to the absolute value of the correlation coefficient being greater than 0.3; the data of the screened environmental indexes is extracted, and the regression analysis is performed by the logistic regression method and the binary coded social state and three behavior type data, respectively, to obtain the regression coefficient β and the OR value, and the influence mode and degree of the environmental index on the social behavior index are determined according to the threshold range of the regression coefficient β and the OR value, and the threshold range is as follows: weak positive influence, 0≤β<0.1 and 1.0≤OR<1.1; medium positive influence, 0.1≤β<0.5 and 1.1≤OR<2.0; strong positive influence, 0.5≤β and 2.0≤OR; weak negative influence, -0.1≤β<0 and 0.9≤OR<1.0; medium negative influence, -0.5≤β<-0.1 and 0.5≤OR<0.9; and strong negative influence, β<-0.5 and OR<0.5.

[0041] Optionally, in step (6), the time-space heat plane of each environmental index, the social behavior heat plane, and the environmental index influence degree ranking table are output in three dimensions by the holographic digital projector and the visualization system.

[0042] Beneficial Effects: Compared with existing technologies, this invention has the following significant advantages: This invention combines the specific characteristics of the target public space and integrates environmental sensing devices and behavioral sensing devices through a sensing device network, achieving precise coupling between physical environment data and social behavior data, thus improving the accuracy of behavioral environment impact analysis and data collection efficiency; This invention integrates temperature and humidity sensors, light sensors, and wind environment sensors into environmental sensing devices, achieving comprehensive measurement of the physical environment, more comprehensively assessing the impact of the physical environment on social behavior, and obtaining the impact rating of each physical environment indicator; This invention analyzes whether an individual is in a social state and the type of social behavior they are engaging in by recognizing social distance, facial orientation, and body movements, thus refining and improving the accuracy of social behavior recognition; This invention constructs a data correction method based on the spatiotemporal characteristics of the data, eliminating the influence of fixed spatial locations and time nodes on social behavior, making the relationship between the physical environment and social behavior clearer and more explicit; This invention calculates the number and location of sensing devices, optimizes device coverage, and improves the comprehensiveness and accuracy of data collection; This invention utilizes HTML container elements to create interactive maps, allowing users to easily select public space areas, improving user experience and participation. Attached Figure Description

[0043] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0044] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0045] like Figure 1 As shown, a method for identifying social behavior and environmental impact in public spaces based on a network of sensing devices includes the following steps:

[0046] S1. Public space area selection: Select public space areas using the interactive map on the web interface, and record the geographical location information of the public space areas based on the user's selection input;

[0047] An interactive map is loaded and displayed on the webpage using an HTML container element. It receives user input on the map, which refers to selecting a public space area on the map. Based on the user's input, the geographic location information of the public space area is recorded. The geographic location information includes the geometric information of all vertex coordinates, the geographic coordinates of the geometric center point, and the geographic area of ​​the public space area.

[0048] S2. Measurement point layout and perception device allocation: determine the number of perception devices and the measurement point layout based on the geographic location information of the public space area in S1, deploy a perception device network including environmental perception devices and behavior perception devices within the public space area, determine the measurement point layout of the environmental perception devices and the behavior perception devices, install the environmental perception devices and the behavior perception devices, test the working state and coverage of the devices, and transmit 24-hour continuous monitoring data to the data center;

[0049] The environmental perception devices integrate temperature and humidity sensors, light sensors, and wind environment sensors and are evenly distributed in the public space area in a 20m x 20m grid; the behavior perception devices integrate a movable power supply and a network camera, and the effective sensing range of the behavior perception devices is a planar area with a radius of 50m and an angle of 80° centered on the device, the boundary and geometric center of the public space area are determined, the public space area is divided into 5m x 5m grid cells, and the number of devices and the layout positions of the devices for full grid cell coverage and minimum overlap are calculated through a genetic algorithm;

[0050] S3. Public space physical environment monitoring: pre-process the environmental data collected by the environmental perception devices, i.e. clean up outliers, aggregate the environmental data to the corresponding spatial grid cells according to time nodes, generate a grid data matrix with time series information, and generate continuous environmental data distribution of the public space area through spatial interpolation method, calculate the average value of environmental data within a unit time interval, and store it in the database in a structured data form;

[0051] The structured data in step S3 includes the grid cell identifier of the data, the timestamp, the environmental indicator type, and the data average value, the environmental indicator type includes the light environment indicator, the wind environment indicator, and the thermal environment indicator, the light environment indicator includes the sunshine duration per minute and per hour, the sunshine duration is calculated according to the time when the direct radiation measured by the light sensor is less than or equal to 120W / ㎡; the wind environment indicator includes the average wind speed and the average wind direction per minute, the average wind speed and the average wind direction per minute are calculated according to the wind speed and the wind direction measured by the wind environment sensor; the thermal environment indicator includes the average temperature and the relative humidity per minute, the average temperature and the relative humidity per minute are calculated according to the temperature and the humidity measured by the temperature and humidity sensor;

[0052] S4. Public space social behavior recognition: real-time process the video data of the behavior perception devices through a convolutional neural network, detect and track individuals in the crowd, obtain individual position coordinates and motion trajectories; eliminate the influence of occlusions and video angles through multi-camera data fusion technology, and merge redundant data; combine social distance, facial orientation, and limb action recognition to determine the social state and social behavior type of individuals, add spatial grid data to the social behavior indicator, and store it in the database in a structured data form;

[0053] The multi-camera data fusion technology in step S4 refers to combining the position coordinates (x c ,y c ,z c ) and the shooting direction (θ x ,θ y ,θ z ) of the camera, the center point coordinates (c x ,c y ) of the individual image in the photographic image, and the focal lengths f x and f y of the x-axis and y-axis on the photographic image plane determined by the vertical angle of the camera and the sensor size of the camera, to construct the intrinsic matrix and the extrinsic matrix [R|T] of the camera, wherein the rotation matrix R = R z (θ z )·R y (θ y )·R x (θ x ), wherein R x , R y , R z are rotation matrices of the rotation angles around the x, y, and z axes, respectively, the translation matrix wherein t x , t y , and t z are the translation amounts of the center point coordinates of the individual image in the photographic image along the x, y, and z axes, respectively, and the projection matrix P = K·[R|T] of the camera is defined, and the individual image coordinates are converted into spatial plane coordinates through the projection matrix; based on the converted spatial plane coordinates, the trajectories of each individual in the camera are recorded to obtain a series of grid coordinate points, the trajectory data of each camera is matched through the timestamp and the spatial position, and the distance of two trajectories within a unified time period is calculated wherein (X i,t ,Y i,t ) represents the coordinate position of individual i at time frame t, and (X j,t ,Y j,t ) represents the coordinate position of individual j at time frame t; the distance threshold is set to 2m, and when d < 2m, the individual similarity is identified through the cross-device pedestrian re-identification ReID (Re-Identification); a ResNet deep learning model is used to extract the feature vector of the individual image captured by each camera, and the feature vector similarity between different individuals within a unified time period under different cameras is calculated through the cosine similarity Cosine Similarity:

[0054]

[0055] Wherein: x·y is the dot product of the eigenvectors, ||x|| and ||y|| are the Euclidean norms of vectors x and y respectively; set the cosine similarity threshold to 0.5, and consider the eigenvectors with cosine similarity > 0.5 as the same object and assign the same identity ID; by associating the trajectories of the same identity ID in different cameras and merging the trajectories matched in each camera, the repeated trajectories are removed to form unified trajectory data; the trajectories matched in different cameras are merged into a continuous trajectory sequence, the planar coordinates, video frame images and identity ID information in each time frame are retained, and the trajectory data including the planar grid coordinates (X t ,Y t ) of each time frame t and the corresponding video image after merging are stored in the database in the form of structured data.

[0056] In step S4, the social state and social behavior type of the individual are judged by combining social distance, face orientation and limb motion recognition. First, the database is sorted, the motion trajectory of the individual is segmented according to the time stamp, and the video frame is extracted at 0.5s time interval; by analyzing the motion trajectory, according to the distance between any two individuals at different time points of the planar coordinate data, for each time stamp t, the Euclidean distance d ij (t) between individual i and individual j is calculated:

[0057]

[0058] Wherein: x i (t), y i (t) are the planar coordinates of individual i at time t, based on the preset social distance threshold D threshold , if d ij (t)≤D threshold and the duration Δt exceeds the threshold, it is judged that there is social interaction between two or more individuals; by positioning the face region through the MTCNN face detection algorithm, the Euler angles (α, β, γ) of the head are calculated through head pose estimation, the facing direction between individuals is matched, and if the difference between the yaw angles β i and β j of two or more individuals is within a certain threshold Δβ threshold , i.e. |β i -β j |≤Δβ threshold , and the relative distance meets the social distance threshold, it is judged that social interaction is being carried out; by extracting the key node coordinates P={p1,p2...p n} in each video frame through the OpenPose human pose estimation algorithm, the motion trajectory and speed of the key limb parts are calculated by analyzing the time sequence of the key nodes, and the movement distance of the key node k between time t and t+Δt:

[0059]

[0060] The extracted feature vector v = {v1, v2...v n} is input into the SVM classification model, and the social distance d ij (t), face orientation β, and limb movement feature v are comprehensively analyzed:

[0061] F = concat(d ij (t), β i , β j , v)

[0062] The random forest model outputs the probability distribution P(class|F), and based on the feature vector F obtained by splicing multiple features, the corresponding recognition is classified into three behavior types of health class, entertainment class, and ordinary social class, and finally the structured data set including individual ID, motion trajectory, social state, social behavior type, time and space position of behavior occurrence, and corresponding motion state detailed information is output;

[0063] The structured data in step S4 includes grid cell identifier, timestamp, behavior index type, and index data value of the data, the behavior index type includes social state and social behavior type, the social state includes two data values of yes and no, and the social behavior type includes three data values of health class, entertainment class, and ordinary social class;

[0064] S5. Behavior environment index influence analysis: based on the structured data of physical environment in S3 and the structured data of social behavior in S4, the influence of fixed spatial position and time node on social behavior is excluded through spatial regression model and time series decomposition, and the social behavior data set only retaining the influence of environmental factors is obtained; the correlation analysis and multiple regression analysis are combined to determine whether each environmental index has influence on social behavior, and the influence mode and degree are determined;

[0065] In S5, the influence of fixed spatial position and time node on social behavior is excluded through spatial regression model and time series decomposition. First, the spatial autocorrelation of the structured social behavior data is calculated, a spatial weight matrix W is constructed for capturing the mutual relationship between spatial positions, wherein the element W ij of the matrix represents the spatial interaction between position i and position j, and a spatial autoregressive model SAR is constructed:

[0066] Y ij = ρWY ij + X ij β + ∈ ij

[0067] Where: Y ijis the social behavior of individual i at location j, ρ is the spatial autocorrelation coefficient, X ij is the environmental index, β is the coefficient vector of independent variables, ∈ ij is the error term;

[0068] The model parameters ρ and β are estimated using maximum likelihood estimation (MLE), and based on the estimated ρ, the spatial autocorrelation part of the social behavior data is calculated:

[0069] Y spatial = ρWY

[0070] The spatial autocorrelation part is removed from the original data to obtain the non-spatial autocorrelation part, i.e., the residual Y residual :

[0071] Y residual = Y - Y spatial - Xβ

[0072] The residual Y residual is the social behavior data after removing spatial autocorrelation and excluding fixed spatial positions;

[0073] Secondly, the structured social behavior data is decomposed into a time series to calculate:

[0074] Y t = T t + S t + R t

[0075] where Y t is the social behavior data, T t is the trend component, S t is the seasonal component, and R t is the residual component; removing the trend and seasonality from the original data gives the residual R t = Y t - T t - S t , which is the social behavior data excluding the influence of time nodes; combining the residual data Y residual from the spatial autocorrelation model with the residual data R t from the time series decomposition gives the comprehensive spatial and temporal adjusted social behavior data set Y adj :

[0076] Y adj = Y residual + R t

[0077] The data set Y adj excludes the influence of fixed spatial positions and time nodes on social behavior, and retains the behavior change characteristics related to the environment;

[0078] S5. Correlation analysis and multiple regression analysis: Determine whether each environmental indicator has an impact on social behavior and the degree of impact. Standardize the environmental indicator data in the structured data from S3, and encode the social status and three behavior types in the social behavior data set Y adj into binary variables, with "0" representing "no" and "1" representing "yes". Calculate the correlation coefficients between the environmental indicator data and the social status data and the three behavior type data through point correlation analysis, and select the environmental indicators that are significantly correlated with social status and behavior type, with a correlation coefficient absolute value greater than 0.3. Extract the data of the selected environmental indicators, and perform regression analysis with the binary coded social status and three behavior type data through logistic regression method to obtain the regression coefficient β and OR value. Determine the impact of the environmental indicator on the social behavior indicator according to the threshold range of the regression coefficient β and OR value, with the threshold range as follows: weak positive impact, 0≤β<0.1 and 1.0≤OR<1.1; moderate positive impact, 0.1≤β<0.5 and 1.1≤OR<2.0; strong positive impact, 0.5≤β and 2.0≤OR; weak negative impact, -0.1≤β<0 and 0.9≤OR<1.0; moderate negative impact, -0.5≤β<-0.1 and 0.5≤OR<0.9; strong negative impact, β<-0.5 and OR<0.5.

[0079] S6. Impact rating and output: Sort the impact of each environmental indicator on social behavior based on the regression analysis results in step S5, and display the time-space heat plane of each environmental indicator, the social behavior heat plane, and the environmental indicator impact degree sorting table in three dimensions through a holographic digital projector and a visualization system.

Claims

1. A method for identifying social behavior and environmental impact in public spaces based on a network of sensing devices, characterized in that, Includes the following steps: (1) Select public space areas using the interactive map on the webpage interface, and record the geographical location information of the public space areas based on the user's selection input; (2) Based on the geographical location information of the public space area in step (1), determine the number of sensing devices and the layout of measurement points, deploy a sensing device network in the public space area, the sensing device network includes environmental sensing devices and behavioral sensing devices, test the working status and coverage of the devices, and transmit continuous monitoring data to the data platform. (3) The physical environment data of the public space area collected by the environmental sensing device is summarized into the corresponding spatial grid unit according to the time node, generating a grid data matrix with time series information, and generating the continuous environmental data distribution of the public space area through spatial interpolation method, calculating the average value of the environmental data within the unit time interval, and storing it in the database in the form of structured data; (4) The video data of the behavior perception device is processed in real time through convolutional neural network to detect and track individuals in the crowd and obtain individual position coordinates and movement trajectories; the influence of occlusion and video angle is eliminated through multi-camera data fusion technology and redundant data is merged; the social status and social behavior type of individuals are judged by combining social distance, facial orientation and body movement recognition, and spatial grid data is added to the social behavior indicators and stored in the database in the form of structured data. (5) Based on the structured data of the physical environment in step (3) and the structured data of social behavior in step (4), the influence of fixed spatial location and time node on social behavior is excluded by spatial regression model and time series decomposition to obtain a social behavior dataset that retains only the influence of environmental factors; combined with correlation analysis and multivariate regression analysis, it is determined whether each environmental indicator has an influence on social behavior, and the mode and degree of influence are determined. (6) Based on the regression analysis results in step (5), the influence of various environmental indicators in public space on social behavior is ranked and displayed in three dimensions.

2. The method for identifying social behavior and environmental impact in public spaces based on a network of sensing devices according to claim 1, characterized in that: Step (1) specifically includes the following steps: Loading and displaying an interactive map on the web page interface through an HTML container element, receiving user selection input on the map, where the selection input refers to selecting a public space area in the map; recording the geographical location information of the public space area based on the user's selection input, where the geographical location information includes the geometric information of all vertex coordinates of the public space area, the geographical coordinates of the geometric center point, and the geographical area.

3. The method for identifying social behavior and environmental impact in public spaces based on a network of sensing devices according to claim 1, characterized in that: In step (2), the environmental sensing device integrates temperature and humidity sensors, light sensors, and wind environment sensors, which are evenly distributed in the public space area in a 20m×20m grid. The behavior sensing device integrates a portable power supply and a network camera. The effective sensing range of the behavior sensing device is a planar area with a radius of 50m and an angle of 80° centered on the device. The boundary and geometric center of the public space area are determined, and the public space area is divided into 5m×5m grid units. The number of devices and their locations that achieve full grid unit coverage and minimum overlap are calculated through a genetic algorithm.

4. The method for identifying social behavior and environmental impact in public spaces based on a network of sensing devices according to claim 1, characterized in that: The structured data in step (3) includes the grid cell identifier, timestamp, environmental index type, and average data value. The environmental index types include light environment index, wind environment index, and thermal environment index. The light environment index includes sunshine hours per minute and per hour, calculated cumulatively based on the time when direct radiation measured by the light sensor is less than or equal to 120W / ㎡. The wind environment index includes average wind speed and average wind direction per minute, calculated based on the wind speed and wind direction measured by the wind environment sensor. The thermal environment index includes average temperature and relative humidity per minute, calculated based on the temperature and humidity measured by the temperature and humidity sensor.

5. The method for identifying social behavior and environmental impact in public spaces based on a network of sensing devices according to claim 1, characterized in that: The multi-camera data fusion technology in step (4) refers to combining the position coordinates (x, y) of the cameras. c ,y c ,z c ) and shooting direction (θ) x ,θ y ,θ z ), the center point coordinates of an individual image within a photographic image (c x ,c y And the focal length f along the x and y axes of the photographic image plane, determined by the camera's vertical angle and sensor size. x and f y Construct the intrinsic parameter matrix of the camera. And the extrinsic parameter matrix [R|T], where the rotation matrix R = R z (θ z )·R y (θ y )·R x (θ x ), where R x R y R z These are the rotation matrices and translation matrices for rotation angles around the x, y, and z axes, respectively. Where t x t y and t z These represent the translations of the center point coordinates of an individual image along the x, y, and z axes in the captured image. The projection matrix P = K·[R|T] of the camera is then defined, converting the individual image coordinates into spatial plane coordinates. Based on these converted spatial plane coordinates, the trajectories of each object in the camera are recorded, resulting in a series of grid coordinate points. The trajectory data from each camera are then matched using timestamps and spatial locations to calculate the distance between the two trajectories within the same time period. Where (X) i,t ,Y i,t (X) represents the coordinate position of individual i at time frame t. j,t ,Y j,t () represents the coordinate position of individual j at time frame t; a distance threshold of 2m is set, and when d < 2m, individual similarity is identified through cross-device pedestrian re-identification (Re-Identification); the ResNet deep learning model is used to extract the feature vector of individual images captured by each camera, and the similarity of feature vectors between different individuals in the same time period under different cameras is calculated by cosine similarity: Where: x·y is the dot product of feature vectors, and ||x|| and ||y|| are the Euclidean norms of vectors x and y, respectively; the cosine similarity threshold is set to 0.5, and feature vectors with cosine similarity > 0.5 are considered the same object and assigned the same identity ID; by associating trajectories with the same identity ID in different cameras and merging the matched trajectories in each camera, duplicate trajectories are removed to form unified trajectory data; the matched trajectories in different cameras are merged into a continuous trajectory sequence, retaining the planar coordinates, video frame images, and identity ID information in each time frame, and the merged sequence includes the planar grid coordinates (X, Y, Y, Y) of each time frame t. t ,Y t The trajectory data of the corresponding video images are stored in the database in the form of structured data.

6. The method for identifying social behavior and environmental impact in public spaces based on a network of sensing devices according to claim 5, characterized in that: In step (4), social distance, facial orientation, and body movements are combined to identify and determine an individual's social status and social behavior type. First, the database is organized, and the individual's movement trajectory is segmented according to the timestamp. Video frames are extracted at 0.5-second time intervals. By analyzing the movement trajectory, the Euclidean distance d between individuals i and j is calculated for each timestamp t based on the distance between any two individuals at different time points in the planar coordinate data. ij (t): Where: x i (t), y i (t) represents the planar coordinates of individual i at time t, based on a preset social distance threshold D. threshold If d ij (t)≤D threshold If the duration Δt exceeds a threshold, then social interaction is determined to exist between two or more individuals. The MTCNN face detection algorithm is used to locate face regions. Head pose estimation is used to calculate the Euler angles (α, β, γ) of the head. The facing directions of individuals are matched; if the yaw angle β of two or more individuals exceeds the threshold, then social interaction is considered. i and β j The difference between them is at a certain threshold Δβ threshold Inside, i.e., |β i -β j |≤Δβ threshold Furthermore, the relative distance meets the social distance threshold, indicating that they are engaging in social interaction; the key node coordinates P = {p1, p2... p...} in each video frame are extracted using the OpenPose human pose estimation algorithm. n By analyzing the time series of key points, the motion trajectory and velocity of critical limb parts are calculated, and the distance moved by key point k between time t and t+Δt is determined. The extracted feature vector v = {v1, v2, ... v} n The input is fed into the SVM classification model, incorporating social distance d. ij A comprehensive analysis was conducted using (t), facial orientation β, and limb movement characteristics v: F=concat(d ij (t),β i ,β j ,v) The random forest model outputs a probability distribution P(class|F). Based on the feature vector F, which is a concatenation of multiple features, the model is used to identify three types of behavior: health and fitness, entertainment and general social interaction. The final output is a structured dataset that includes individual ID, movement trajectory, social status, social behavior type, time and space of the behavior, and detailed information about the corresponding movement status.

7. The method for identifying social behavior and environmental impact in public spaces based on a network of sensing devices according to claim 6, characterized in that: The structured data in step (4) includes the grid cell identifier, timestamp, behavior indicator type and indicator data value. The behavior indicator type includes social status and social behavior type. The social status includes two data values: yes and no. The social behavior type includes three data values: health and fitness, entertainment and general social.

8. The method for identifying social behavior and environmental impact in public spaces based on a network of sensing devices according to claim 1, characterized in that: In step (5), the influence of fixed spatial location and time node on social behavior is eliminated through spatial regression model and time series decomposition. First, spatial autocorrelation is calculated on the structured social behavior data to construct a spatial weight matrix W to capture the relationship between spatial locations, where the elements of the matrix W ij Represent the spatial interaction between positions i and j, and construct a spatial autoregressive (SAR) model: Y ij =ρWY ij +X ij β+∈ ij Where: Y ij ρ is the social behavior of individual i at location j, ρ is the spatial autocorrelation coefficient, and X is the spatial autocorrelation coefficient. ij It is an environmental indicator, β is the coefficient vector of the independent variable, ∈ ij It is an error term; Maximum likelihood estimation (MLE) is used to estimate the model parameters ρ and β. Based on the estimated ρ, the spatial autocorrelation component of the social behavior data is calculated. Y spatial =ρWY The spatial autocorrelation component is removed from the original data to obtain the non-spatial autocorrelation component, i.e., the residual Y. residual : Y residual =Y-Y spatial -Xβ Residual Y residual That is, after removing spatial autocorrelation, social behavior data at fixed spatial locations are excluded; Secondly, time-series decomposition calculations are performed on the structured social behavior data: Y t =T t +S t +R t Among them, Y t For social behavior data, T t As a trend component, S t As a seasonal component, R t The residual components are obtained by removing trends and seasonality from the original data. t =Y t -T t -S t This refers to social behavior data that has excluded the influence of time points; the residual data Y obtained from the spatial autocorrelation model residual The residual data R obtained from time series decomposition t By combining these components, we obtain the social behavior dataset Y, which is adjusted for both spatial and temporal factors. adj : Y adj =Y residual +R t Dataset Y adj It eliminates the influence of fixed spatial location and time point on social behavior, and retains the characteristics of behavior changes related to the environment.

9. The method for identifying social behavior and environmental impact in public spaces based on a network of sensing devices according to claim 8, characterized in that: In step (5), combining correlation analysis and multiple regression analysis to determine whether various environmental indicators have an impact on social behavior, and to determine the mode and extent of the impact, refers to standardizing the environmental indicator data in the structured data of the physical environment in step (3), and then standardizing the social behavior dataset Y. adj The social status and three behavioral types data were coded as binary variables, where "0" represents "no" and "1" represents "yes". Correlation coefficients between environmental indicator data and social status data, and between the three behavioral types data, were calculated using point-binary correlation analysis. Environmental indicators significantly correlated with social status and behavioral types were selected, with a correlation coefficient absolute value greater than 0.3 indicating significant correlation. The selected environmental indicator data were then extracted and subjected to logistic regression analysis with the binary-coded social status and three behavioral types data to obtain the regression coefficient β and OR value. The influence of an environmental indicator on a social behavior indicator is determined based on its threshold range, which is as follows: weak positive influence, 0 ≤ β < 0.1 and 1.0 ≤ OR < 1.1; moderate positive influence, 0.1 ≤ β < 0.5 and 1.1 ≤ OR < 2.0; strong positive influence, 0.5 ≤ β and 2.0 ≤ OR; weak negative influence, -0.1 ≤ β < 0 and 0.9 ≤ OR < 1.0; moderate negative influence, -0.5 ≤ β < -0.1 and 0.5 ≤ OR < 0.9; strong negative influence, β < -0.5 and OR < 0.

5.

10. The method for identifying social behavior and environmental impact in public spaces based on a network of sensing devices according to claim 1, characterized in that: In step (6), a holographic digital projector and a visualization system are used to display and output the time-space thermal plane of various environmental indicators, the social behavior thermal plane, and the ranking table of the degree of influence of environmental indicators in three dimensions.

Images